I am doing a competition grow assay where GFP+ keratinocytes and mCherry+ kerationcytes are plated together, allowed to grow and passed multiple times. At each passage, cells are plated onto a coverslip, fixed and stained with DAPI. See the attached image for a representative image from this experiment.
I have figured out how to get Image J to automatically count DAPI+ cells but I'm having issues getting Image J to accurately count the GFP+ (or mCherry) cells. One major issues is that there is not a well defined separation between many of the neighboring cells. I've tried many parameters within the watershed function but the counts are significantly off from the counts I got manually. I believe I need a function where cells are counted only if they are both GFP and dapi positive. (I know my mCherry signal is very weak and there is a lot of background noise. I'm okay with counting all the dapi+ cells then counting dapi+gfp and subtracting those numbers to get my mCherry count.)
I have about 300 images to process, so as much as I can automate the more sanity I can keep!
Thank you for you help and time.
For what I understand you have diffuse EGFP and Cherry. Why don't you just use a trick and create a mask using your DAPI channel? By doing this you would get an image with nuclei in green or in red and you can count them easily. If your aim is just to have a count that should work. I am downloading your image and give it a try, stay tuned!
Here is a rough try. It does look countable to me, of course you have to play a bit with the threshold to see which works best for your data.
I first thresholded the DAPI channel and converted into a 0-1 mask. You apply/multiply on your egfp or cherry image to obtain green or red nuclei. I suggest you first threshold your egfp / cherry image.
Thank you Francesco, Yes, I just need a count. I spent days reading, try to get colocalization to work for my purpose. I'll definitely play around with masking feature today. I'll let you know how it goes.
Hmm, I wonder if CellProfiler (http://cellprofiler.org/) may be helpful in trying to automate these counts? Hopefully setting up the pipeline is less tedious than manually counting!
Thank you Robert for your reply! I got ImageJ to work well for my purpose.
These are the abbreviated steps I'm using:
Split channels
Count dapi
Subtract the red image from green.
Then using the AND function in the image calculator, merge the subtracted green image AND the masked dapi image. This will turn some of the dapi spots green.
count the green dapi.
And for the reverse, subtract the green image from red and repeat the AND function with the red subtracted image. Count dapi.
Thank you Francesco for you suggestion. Although I couldn't get the multiply function to work, your suggestion got me to the image calculator where I found the solution.
- Badges
- Science topic
- Similar topics
- Computer Science
- Images